betaFeedback

WorksPhysics and Astronomy

Thermoelectric and Magnetic Properties in Doped Fe2_2VAl within a Bipolar Random Anderson Model

Abstract

We investigate the thermoelectric and magnetic properties of Si-substituted nn-type and Ti-substituted pp-type Heusler alloy Fe2_2VAl using the bipolar random Anderson model, which has been introduced recently to study antisite-defect effects associated with the sign change of the Seebeck coefficient in thermally quenched Fe2_2VAl. Based on the electronic states of both nn-type and pp-type compounds, with the rigid-band shift of the Fermi energy and a temperature-dependent scattering rate taken into account, we elucidate how antisite defects simultaneously influence thermoelectric transport and local magnetic moments. We find that the magnetic moments are enhanced in nn-type Fe2_2VAl, whereas they are suppressed in pp-type Fe2_2VAl compared with the undoped compound. These contrasting magnetic responses highlight the impact of antisite spin polarization on thermoelectric properties and demonstrate the crucial role of antisite defects in realizing magneto-thermoelectric functionalities in Heusler-type alloys.

Introduction

The Heusler-type intermetallic Fe2VAl\mathrm{Fe}_2\mathrm{VAl}-based compounds are known as promising candidates for thermoelectric materials.1) Stoichiometric Fe2VAl\mathrm{Fe}_2\mathrm{VAl} is nonmagnetic semimetal and the valence-band maximum is located at the Γ\Gamma point, while the conduction-band minimum lies at the X point, with a small band overlap between them,2–5) resulting in a pseudogap around the Fermi level εF\varepsilon_{\mathrm{F}}.6, 7) This subtle electronic structure is the origin of the diverse thermoelectric properties induced by element substitution and quenching processes. In Fe2VAl\mathrm{Fe}_2\mathrm{VAl}, the resistivity ρ\rho increases as the temperature is lowered and the Seebeck coefficient SS exhibits positive values over a wide temperature range,8) indicating that holes are the majority carriers. Upon Si substitution, in Fe2VAl1−xSix\mathrm{Fe}_2\mathrm{VAl}_{1-x}\mathrm{Si}_x, however, ρ\rho decreases as temperature is lowered and SS changes dramatically from positive to negative (nn-type) even for small Si concentrations of x=0.02x=0.02 and 0.030.03.9, 10) This strong sensitivity primarily originates from the introduction of electron carriers, which shifts εF\varepsilon_{\mathrm{F}} to higher energies within the pseudogap. On the other hand in the pp-type Fe2VAl\mathrm{Fe}_2\mathrm{VAl}, Fe2V1−xTixAl\mathrm{Fe}_2\mathrm{V}_{1-x}\mathrm{Ti}_x\mathrm{Al}11–13) realized by substituting Ti for V resulting in holes in the valence band, the resistivity becomes metallic at low temperatures9, 10) similar to nn-type Fe2VAl\mathrm{Fe}_2\mathrm{VAl}.9, 10) The SS remains positive and its magnitude increases with increasing Ti concentration up to x=0.03x=0.03, and then decreases.11) This pp-type behavior has also been observed in Mn-substituted Fe2V1−xMnxAl\mathrm{Fe}_2\mathrm{V}_{1-x}\mathrm{Mn}_x\mathrm{Al}.14) As for understanding these transport features in doped Fe2VAl\mathrm{Fe}_2\mathrm{VAl}, a phenomenological two-band model has been shown to qualitatively account for the temperature dependence of SS in Fe2VAl0.9Si0.1\mathrm{Fe}_2\mathrm{VAl}_{0.9}\mathrm{Si}_{0.1}15) and evaluations of SS in both nn-type and pp-type Fe2VAl\mathrm{Fe}_2\mathrm{VAl} have also been performed using the linearized Boltzmann transport equation with a constant relaxation time combined with first-principles electronic-structure calculations.16)

Besides the transport properties, magnetic properties have a diversity. In contrast to nonmagnetic Fe2VAl\mathrm{Fe}_2\mathrm{VAl}, ferromagnetic order has been observed at low temperatures in Si-substituted nn-type Fe2VAl\mathrm{Fe}_2\mathrm{VAl}17–19) which has been regarded as a weak itinerant ferromagnet,19) since the field-induced magnetic moment is small and weakens with increasing temperature. Furthermore, quenching Fe2VAl0.9Si0.1\mathrm{Fe}_2\mathrm{VAl}_{0.9}\mathrm{Si}_{0.1} from high temperatures generates a magnetic moment an order of magnitude larger than that of the unquenched sample.20) Quenched undoped Fe2VAl\mathrm{Fe}_2\mathrm{VAl} also induces a magnetic moment,21) but its magnitude is only about one-third of that observed in quenched nn-type Fe2VAl0.9Si0.1\mathrm{Fe}_2\mathrm{VAl}_{0.9}\mathrm{Si}_{0.1}20) highlighting the crucial role of electron doping. Since thermal quenching introduces antisite defects, it is reasonable to attribute such large moments to, for example, V antisites on the Fe sublattice (VFe\mathrm{V}_{\mathrm{Fe}}) and Fe antisites on the V sublattice (FeV\mathrm{Fe}_{\mathrm{V}}). Hence, magnetic properties are seen to be strongly affected by antisite defects.

The introduction of antisites also causes remarkable changes in SS: in Fe2VAl\mathrm{Fe}_2\mathrm{VAl}, the sign of SS becomes negative,21) whereas in nn-type Fe2VAl0.9Si0.1\mathrm{Fe}_2\mathrm{VAl}_{0.9}\mathrm{Si}_{0.1}, the magnitude of SS is reduced.20) A reduction in SS in Fe2VAl1−xSix\mathrm{Fe}_2\mathrm{VAl}_{1-x}\mathrm{Si}_x has also been reported in thin films22) and annealed samples. Regarding pp-type Fe2VAl\mathrm{Fe}_2\mathrm{VAl}, no systematic study has clarified the role of the interplay between antisite defects and their spin polarization. It is therefore necessary to provide a theoretical framework that elucidates the microscopic mechanism of roles played by antisite defects. Hence, systematic understanding based on microscopic analyses in terms of model Hamiltonians is needed to clarify how carrier doping affects the thermoelectric and magnetic properties.

In this paper, we theoretically investigate the thermoelectric properties of doped Fe2VAl\mathrm{Fe}_2\mathrm{VAl} and the spin polarization associated with antisite defects. We first confirm that a simple rigid-band picture, in which εF\varepsilon_{\mathrm{F}} shifts upon carrier doping, captures the overall features of thermoelectric behavior of doped Fe2VAl\mathrm{Fe}_2\mathrm{VAl}. To incorporate antisites into our theory, we employ an extended version of the single-impurity Anderson model,24) the bipolar random Anderson model (BPRAM),25) which has been introduced recently. In this model, VFe\mathrm{V}_{\mathrm{Fe}} and FeV\mathrm{Fe}_{\mathrm{V}} antisites are treated as randomly distributed impurities in the valence and conduction bands, respectively, and induce carrier redistribution between the two bands while conserving the total electron number. This BPRAM framework has successfully explained the sign change of SS caused by antisite defects in Fe₂VAl.25) In nn-type Fe₂VAl, we find that both VFe_{\mathrm{Fe}} and FeV_{\mathrm{V}} antisites are spin-polarized, in contrast to the undoped case where only VFe_{\mathrm{Fe}} exhibits spin polarization.25) This leads to a magnetic moment approximately twice that of the antisites in undoped Fe₂VAl, when the antisite concentration is fixed. The Seebeck coefficient SS decreases upon antisite introduction, consistent with experimental observations.20) In pp-type Fe₂VAl, by contrast, antisite defects lose their magnetic moments because their levels shift above εF\varepsilon_{\mathrm{F}}, resulting in only a small change in SS. The interplay between thermoelectric and magnetic properties is seen to be governed by the spectral conductivity, which is strongly influenced by spin-polarized antisite defects. These findings provide a new guideline for controlling spin-polarized antisite defects in doped Fe₂VAl and demonstrate the crucial role of antisite defects in realizing magneto-thermoelectric functionalities in Heusler-type alloys.

This paper is organized as follows. In Sect. 2, we present calculated results for the density of states, spectral conductivity, resistivity, and Seebeck coefficient in doped Fe₂VAl. In Sect. 3, we introduce the BPRAM and show the electronic states and thermoelectric properties in the presence of antisite defects. Finally, Sect. 4 discusses the implications of the proposed spin polarization of antisite defects and provides a summary.

Thermoelectric Properties in doped Fe₂VAl

First-principles electronic structure calculations based on the local density approximation or generalized gradient approximation for Fe₂VAl have shown a negative band gap, Eg<0E_{\mathrm{g}}<0, in which the valence-band maximum at Γ\Gamma lies higher in energy than the conduction-band minimum at X.2–5) A two-band model with Eg=−0.03E_{\mathrm{g}}=-0.03 eV and the Fermi energy εF\varepsilon_{\mathrm{F}} located at the bottom of the conduction band leads to results consistent with the experimentally observed SS and ρ\rho.25) Figure 1(a) illustrates this model.

Image not extracted

Figure 1. Schematic illustration of the electronic states in Fe₂VAl. (a) undoped Fe₂VAl, (b) nn-type and (c) pp-type Fe₂VAl. The valence band around the Γ\Gamma point is predominantly composed of Fe 3dd orbitals with a degeneracy of νv=2\nu^{\mathrm{v}}=2, whereas the conduction band near the X point mainly consists of V 3dd orbitals with a degeneracy of νc=1\nu^{\mathrm{c}}=1. The top of the valence band is set as the energy reference. The Fermi energy εF\varepsilon_{\mathrm{F}} is indicated by horizontal dashed lines. The energy gap EgE_{\mathrm{g}}, defined as the energy difference between the valence-band top and the conduction-band bottom, is negative (Eg<0E_{\mathrm{g}}<0) in (a), whereas Eg=0E_{\mathrm{g}}=0 in (b) and (c) (see text).

According to our previous work,25) we assume an energy dispersion approximated in terms of effective masses. The density of states (DOS) per spin ss for the valence (v) and conduction (c) bands is given by

Dsα(ε)=−1πIm⁡gsα(ε)(1)D_{s}^{\alpha}(\varepsilon)=-\frac{1}{\pi}\operatorname{Im}g_{s}^{\alpha}(\varepsilon) \tag*{(1)}

with α=v\alpha=\mathrm{v} or c\mathrm{c}. The local Green’s function gsα(ε)g_{s}^{\alpha}(\varepsilon) is given by

gsα(ε)=[zsα+iτsα(1+zsα)]ναWα,(2)g_{s}^{\alpha}(\varepsilon)=\frac{[z_{s}^{\alpha}+i\tau_{s}^{\alpha}(1+z_{s}^{\alpha})]^{\nu^{\alpha}}}{W^{\alpha}}, \tag*{(2)}

where WαW^{\alpha} is the half bandwidth with Wv(c)=1.52W^{\mathrm{v(c)}}=1.52 (2.0) eV and να\nu^{\alpha} is the band degeneracy with νv(c)=2\nu^{\mathrm{v(c)}}=2 (1).25) Here, τsα=[(1−zsα)/(1+zsα)]1/2\tau_{s}^{\alpha}=[(1-z_{s}^{\alpha})/(1+z_{s}^{\alpha})]^{1/2} and zsα=[z~+Σsα(z~)]/Wαz_{s}^{\alpha}=[\tilde{z}+\Sigma_{s}^{\alpha}(\tilde{z})]/W^{\alpha}, where the self-energy Σsα(z~)\Sigma_{s}^{\alpha}(\tilde{z}) arises from hybridization with antisite defects, as described below. By introducing an intrinsic scattering rate δ0\delta_{0} for a system without antisite defects, the complex energy is defined as z~=ε~+iδ0\tilde{z}=\tilde{\varepsilon}+i\delta_{0} with ε~=ε+Wv\tilde{\varepsilon}=\varepsilon+W^{\mathrm{v}} for the valence band and ε~=ε−Wc−Eg\tilde{\varepsilon}=\varepsilon-W^{\mathrm{c}}-E_{\mathrm{g}} for the conduction band, where the top of the valence band is taken as the origin of energy (see Figure 1).

The spectral conductivity σsα(ε)\sigma_{s}^{\alpha}(\varepsilon), which governs electrical and heat currents, is given by26)

σsα(ε)=(e2π/3)(Fα/Wα)2Lsα(ε~)(3)\sigma_{s}^{\alpha}(\varepsilon)=\left(e^{2}\pi/3\right)(F^{\alpha}/W^{\alpha})^{2}L_{s}^{\alpha}(\tilde{\varepsilon}) \tag*{(3)}

with

Lsα(ε~)=(Im⁡zsα)2−12Re⁡τsα(1+zsα)[1−(zsα)2Im⁡zsα+3izsα],(4)L_{s}^{\alpha}(\tilde{\varepsilon})=(\operatorname{Im}z_{s}^{\alpha})^{2}-\frac{1}{2}\operatorname{Re}\tau_{s}^{\alpha}(1+z_{s}^{\alpha})\left[\frac{1-(z_{s}^{\alpha})^{2}}{\operatorname{Im}z_{s}^{\alpha}}+3i z_{s}^{\alpha}\right], \tag*{(4)}

where ee is the elementary charge and FαF^{\alpha} was determined to reproduces thermoelectric properties in Fe₂VAl: Fv=7.0F^{\mathrm{v}}=7.0 eV and Fc=11.2F^{\mathrm{c}}=11.2 eV.25)

Since the valence-band top and conduction-band bottom are located at different momenta, we assume that there is no direct interplay between the two bands. Under this assumption, the total spectral conductivity is given by σst(ε)=σsv(ε~)+σsc(ε~)\sigma_{s}^{\mathrm{t}}(\varepsilon)=\sigma_{s}^{\mathrm{v}}(\tilde{\varepsilon})+\sigma_{s}^{\mathrm{c}}(\tilde{\varepsilon}). In terms of σst(ε)\sigma_{s}^{\mathrm{t}}(\varepsilon), the electrical conductivity L11L_{11} and thermoelectric conductivity L12L_{12}, describing the electrical current density j=L11E+L12(−∇T)\mathbf{j}=L_{11}\mathbf{E}+L_{12}(-\nabla T) within the linear response theory under an electric field E\mathbf{E} and a temperature gradient ∇T\nabla T,27) are given by the Sommerfeld-Bethe relationship,

L11=∫−∞∞(−∂f(ε)∂ε)∑sσst(ε)dε(5)L_{11}=\int_{-\infty}^{\infty}\left(-\frac{\partial f(\varepsilon)}{\partial\varepsilon}\right)\sum_{s}\sigma_{s}^{\mathrm{t}}(\varepsilon)d\varepsilon \tag*{(5)}

and

L12=−1e∫−∞∞(−∂f(ε)∂ε)(ε−μ)∑sσstdε.(6)L_{12}=-\frac{1}{e}\int_{-\infty}^{\infty}\left(-\frac{\partial f(\varepsilon)}{\partial\varepsilon}\right)(\varepsilon-\mu)\sum_{s}\sigma_{s}^{\mathrm{t}}d\varepsilon. \tag*{(6)}

Here, f(ε)f(\varepsilon) is the Fermi distribution function, f(ε)=1/(exp⁡((ε−μ)/(kBT))+1)f(\varepsilon)=1/(\exp((\varepsilon-\mu)/(k_{\mathrm{B}}T))+1), where kBk_{\mathrm{B}} is the Boltzmann constant. The chemical potential μ\mu is a temperature-dependent quantity, μ=μ(T)\mu=\mu(T), determined so as to give the total electron density n=∑s∫−∞∞f(ε)Dst(ε)dεn=\sum_{s}\int_{-\infty}^{\infty}f(\varepsilon)D_{s}^{\mathrm{t}}(\varepsilon)d\varepsilon with the total DOS per spin Dst(ε)=Dsv(ε)+Dsc(ε)D_s^t(\varepsilon) = D_s^v(\varepsilon) + D_s^c(\varepsilon). The ρ\rho and SS are given by ρ=1/L11\rho= 1/L_{11} and S=L12/L11S = L_{12}/L_{11}, respectively. In general, L12L_{12} contains contributions from both electrons and phonons;[28] however, the phonon-drag contribution is expected to be small in the present system and is neglected in the following.

Si substitution is expected to modify the underlying electronic states. One obvious effect is an increase in electron carriers due to the replacement of Al by Si, which results in a shift of εF\varepsilon_{\mathrm{F}} toward higher energies. Another possible effect is a change of EgE_g. Indeed, an examination of first-principles results for x=0.1x = 0.1[15] reveals a very small increase in EgE_g on the order of 10 meV. Such an enhancement of EgE_g has also been pointed out in a phenomenological two-band analysis of SS.[15] Based on these observations, we take Eg=0E_g = 0 as shown in Figure 1(b). Ti-substituted Fe2_2VAl, on the other hand, introduces hole carriers into the valence band, resulting in a shift of εF\varepsilon_{\mathrm{F}} toward lower energies. At the same time, a upward shift of the V-dominated conduction band is expected, since V is replaced by Ti that has a higher 3d3d energy level. This modification changes EgE_g, and we take Eg=0E_g = 0, as shown in Figure 1(c).

Both n-type Fe2_2VAl1−x_{1-x}Six_x and p-type Fe2_2V1−x_{1-x}Tix_xAl exhibit metallic behavior in ρ\rho at low temperatures for x∼0.1x \sim0.1.[9, 11] This indicates the presence of a temperature-dependent scattering rate arising from electron-phonon interactions. Accordingly, we incorporate a linear temperature dependence into the

scattering rate for the spectral conductivity by replacing δ0\delta_0 in Lsα(ε~)L_s^\alpha(\tilde{\varepsilon}) with δ0+γe−pT\delta_0 + \gamma_{\mathrm{e-p}}T. The parameters δ0\delta_0 and γe−p\gamma_{\mathrm{e-p}}, together with εF\varepsilon_{\mathrm{F}}, are determined so as to reproduce the experimentally observed ρ\rho for x∼0.1x \sim0.1.[9, 11] We obtain δ0=0.1\delta_0 = 0.1 meV for both n-type and p-type Fe2_2VAl. The value of γe−p\gamma_{\mathrm{e-p}} is taken to be 1.38×10−61.38 \times10^{-6} eV/K for the conduction (valence) band and 0.63×10−60.63 \times10^{-6} eV/K for the valence (conduction) band in n-type (p-type) Fe2_2VAl, where the band hosting the carriers is assumed to have a larger γe−p\gamma_{\mathrm{e-p}}. These values of γe−p\gamma_{\mathrm{e-p}} are comparable to those reported for Ge0.87_{0.87}Y0.02_{0.02}Sb0.10_{0.10}Ag0.01_{0.01}Te.[29] Finally, the effects of magnetic ordering reported experimentally[12, 17, 18, 19] are not considered in the present study, since they are expected to become significant only at low temperatures. In addition, possible contributions from magnetic fluctuations to the thermoelectric properties are beyond the scope of the present work and are left for future investigation.

Figure 2(a) shows the total DOS in n-type Fe2_2VAl. The vertical line indicates the position of εF\varepsilon_{\mathrm{F}}. The zero-temperature spectral conductivity is shown in the inset. Figure 2(b) and Figure 2(c) present the calculated ρ\rho and SS, respectively, together with experimental data for x=0.1x = 0.1 in Fe2_2VAl1−x_{1-x}Six_x.[9] The calculated results (red solid lines) qualitatively coincide with the experimental data. The change from positive SS in Fe2_2VAl (black dash-dotted line) to negative values originates from the shift of εF\varepsilon_{\mathrm{F}} into the conduction band. When εF\varepsilon_{\mathrm{F}} decreases from 0.08 eV, corresponding to a decrease in xx, the ρ\rho increases and the peak in SS shifts to lower temperatures (not shown). This trend is qualitatively consistent with experimental observations.[9, 10]

Image not extracted

Figure 2. (Color online) Electronic states and transport properties in n-type Fe2_2VAl. (a) Total DOS per spin. The energy zero is set at the top of the valence band. The inset shows the zero-temperature total spectral conductivity σst\sigma_s^t per spin. The vertical line denotes the position of the Fermi energy εF\varepsilon_{\mathrm{F}}. (b) Temperature dependence of the resistivity ρ\rho. (c) Temperature dependence of the Seebeck coefficient SS. The red lines represent calculated results. The black dotted lines correspond to experimental data for Fe2_2VAl1−x_{1-x}Six_x with x=0.1x = 0.1[9]. For comparison, the calculated results for Fe2_2VAl$[25] are shown by black dash-dotted lines in (b) and (c).

Figure 3(a) shows the DOS and the zero-temperature spectral conductivity for pp-type Fe2_2VAl. Figures 3(b) and 3(c) present the calculated ρ\rho and SS, respectively, together with experimental data for Fe2_2V1−x_{1-x}Tix_xAl.[11] It is seen that the calculated ρ\rho (red solid lines) is located between the experimental x=0.05x=0.05 and x=0.2x=0.2 data, while the calculated SS is smaller than that in undoped Fe2_2VAl at low temperatures below T∼200T \sim200 K and increases with increasing TT toward T=800T=800 K. It is of interest to check experimentally the latter trend at high temperatures.

Doped Fe2_2VAl with antisite defects

Thermal quenching of Fe2_2VAl-based compounds induces antisite defects.[20, 21, 23] Among these defects, we focus on FeV_{\mathrm{V}} and VFe_{\mathrm{Fe}} antisites, which affect the underlying electronic states in the valence V 3d3d and conduction Fe 3d3d band, respectively. An FeV_{\mathrm{V}} (VFe_{\mathrm{Fe}}) antisite can be regarded as an impurity in the valence (conduction) band, which is described by the Anderson model.[24] However, since such antisites are randomly distributed within the host sublattice, their effects are more appropriately described by an Anderson-type model with randomly distributed impurities. In addition, charge transfer between an antisite defect and a host band may occur. Assuming that FeV_{\mathrm{V}} and VFe_{\mathrm{Fe}} are present in equal concentrations, charge conservation requires that the number of electrons transferred from FeV_{\mathrm{V}} antisites to the conduction band equals the number of holes transferred from VFe_{\mathrm{Fe}} antisites to the valence band. This charge redistribution can give rise to bipolar effects and substantially reduce the energy gap EgE_{\mathrm{g}} making it strongly negative.

To account for this phenomenon, we introduce the BPRAM,[25] whose Hamiltonian is given by

HRAMα=∑k,sεkαcksα†cksα+∑i(εdαniα+Udαni↑αni↓α)+Vα∑k,s,i(disα†cksα+H.c.)(7)H_{\mathrm{RAM}}^\alpha= \sum_{\mathbf{k},s} \varepsilon_{\mathbf{k}}^\alpha c_{\mathbf{k}s}^{\alpha\dagger}c_{\mathbf{k}s}^\alpha+ \sum_i \left(\varepsilon_d^\alpha n_i^\alpha+ U_d^\alpha n_{i\uparrow}^\alpha n_{i\downarrow}^\alpha\right) + V^\alpha\sum_{\mathbf{k},s,i}\left(d_{is}^{\alpha\dagger}c_{\mathbf{k}s}^\alpha+ \mathrm{H.c.}\right) \tag*{(7)}

where cksα†c_{\mathbf{k}s}^{\alpha\dagger} is the creation operator for a mobile electron with momentum k\mathbf{k}, spin ss, and dispersion εkα\varepsilon_{\mathbf{k}}^\alpha, disα†d_{is}^{\alpha\dagger} is the creation operator for a localized electron at an antisite ii, and niα=ni↑α+ni↓αn_i^\alpha=n_{i\uparrow}^\alpha+n_{i\downarrow}^\alpha with nisα=disα†disαn_{is}^\alpha=d_{is}^{\alpha\dagger}d_{is}^\alpha. εdα\varepsilon_d^\alpha is the antisite energy level, UdαU_d^\alpha is on-site Coulomb interaction at the antisites, and VαV^\alpha represents the hybridization between mobile and antisite electrons.

Electrons in the valence and conduction bands are scattered by randomly distributed antisite defects via the hybridization term in Eq. (7). This scattering induces a k\mathbf{k}-independent but spin-dependent self-energy for the electrons, denoted by Σσα(ε)\Sigma_\sigma^\alpha(\varepsilon). By applying a single-site approximation for the antisite defects and employing a self-consistent TT-matrix approximation for the self-energy, we obtain the following self-consistent equation,[25]

Σσα(ε)=ci(Vα)2Gsd,α(ε)1−(Vα)2Gsd,α(ε)gsα(ε)(8)\Sigma_\sigma^\alpha(\varepsilon)=c_i\frac{(V^\alpha)^2G_s^{d,\alpha}(\varepsilon)}{1-(V^\alpha)^2G_s^{d,\alpha}(\varepsilon)g_s^\alpha(\varepsilon)} \tag*{(8)}

where cic_i denotes the antisite concentration. Here, Gsd,α(ε)G_s^{d,\alpha}(\varepsilon) is the Green’s function for antisite states, which is written as Gsd,α(ε)=[ε−εsd,α−Σsd,α(ε)]−1G_s^{d,\alpha}(\varepsilon)=[\varepsilon-\varepsilon_s^{d,\alpha}-\Sigma_s^{d,\alpha}(\varepsilon)]^{-1} with Σsd,α(ε)=(Vα)2gα(ε)\Sigma_s^{d,\alpha}(\varepsilon)=(V^\alpha)^2g^\alpha(\varepsilon). The effective antisite energy level is given by εsd,α=εdα+Udα⟨n−sα⟩\varepsilon_s^{d,\alpha}=\varepsilon_d^\alpha+U_d^\alpha\langle n_{-s}^\alpha\rangle, where the spin-dependent antisite occupation number is defined as ⟨nsα⟩=−1π∫−∞εFIm⁡Gsd,α(ε)dε\langle n_s^\alpha\rangle=-\frac{1}{\pi}\int_{-\infty}^{\varepsilon_{\mathrm{F}}}\operatorname{Im}G_s^{d,\alpha}(\varepsilon)d\varepsilon.

For the antisite concentration, we assume ci=0.01c_i=0.01, as in our previous work on Fe2_2VAl.[25] To satisfy the bipolar condition arising from charge transfer together with carrier doping introduced by Si and Ti substitutions, we set the energy gap and Fermi energy to Eg=−0.232E_{\mathrm{g}}=-0.232 eV and εF=−0.062\varepsilon_{\mathrm{F}}=-0.062 eV for nn-type Fe2_2VAl, and Eg=−0.28E_{\mathrm{g}}=-0.28 eV and εF=−0.149\varepsilon_{\mathrm{F}}=-0.149 eV for pp-type Fe2_2VAl. For the antisite energy levels, we assume that their energy offsets from the top of the valence band (for εdv\varepsilon_d^\mathrm{v}) and from the bottom of the conduction band (for εdc\varepsilon_d^\mathrm{c}) remain constant and independent of carrier doping. We set this offset to 0.16 eV. This yields εdv=−0.16\varepsilon_d^\mathrm{v}=-0.16 eV, while εdc=−0.072\varepsilon_d^\mathrm{c}=-0.072 eV for nn-type Fe2_2VAl and εdc=−0.12\varepsilon_d^\mathrm{c}=-0.12 eV for pp-type Fe2_2VAl. Since both εdv\varepsilon_d^\mathrm{v} and εdc\varepsilon_d^\mathrm{c} lie below εF\varepsilon_{\mathrm{F}} in nn-type Fe2_2VAl, we expect spin polarization at both VFe_{\mathrm{Fe}} and FeV_{\mathrm{V}} antisites, using Udc=Udv=0.2U_d^\mathrm{c}=U_d^\mathrm{v}=0.2 eV and Vv=0.1V^\mathrm{v}=0.1 eV as in our previous study.[25] In contrast, no spin polarization is expected for antisite defects in pp-type Fe2_2VAl because both εdv\varepsilon_d^\mathrm{v} and εdc\varepsilon_d^\mathrm{c} lie above εF\varepsilon_{\mathrm{F}}. These expectations are indeed confirmed in our calculations, as shown below. We note that these parameters are chosen as representative values, and our conclusions presented below are robust against small variations in these parameters.

We self-consistently determine the self-energies in Eq. (8) and calculate ρ\rho and SS to examine the effect of antisite defects on the thermoelectric properties of nn-type and pp-type Fe2_2VAl. The red solid lines in Figure 4(a) and Figure 4(b) show the calculated ρ\rho and SS, respectively, for nn-type Fe2_2VAl. The resistivity ρ\rho is reduced compared with the value obtained in the absence of antisites and becomes metallic up to T=800T=800 K. These changes are consistent with the experimental data for quenched Fe2_2VAl0.9_{0.9}Si0.1_{0.1},[20] shown by the blue dotted line. This reduction in ρ\rho originates from an enhancement of the spectral conductivity due to the increased DOS near εF\varepsilon_{\mathrm{F}}.

Image not extracted

Figure 4. (Color online) Temperature dependence of (a) the calculated resistivity ρ\rho and (b) the Seebeck coefficient SS in the BPRAM for nn-type Fe2_2VAl with antisite defects (red solid lines). The blue dashed lines correspond to experimental data for undoped Fe2_2VAl0.9_{0.9}Si0.1_{0.1} containing antisites.[20] For comparison, the results without antisite defects, identical to the red solid lines in Figs. 2(b) and 2(c), are shown by black dashed lines.

To clarify this point, we show the spin-resolved electronic states in the BPRAM in Figure 5. Figure 5(a) and Figure 5(b) display the DOS of the band electrons for each spin. Because of charge transfer, both the valence and conduction bands contribute to the DOS near εF\varepsilon_{\mathrm{F}}, resulting in an enhancement compared with the case without antisites [see Figure 2(a)]. As a result, the total spectral conductivity σ↑t+σ↓t\sigma_{\uparrow}^{t}+\sigma_{\downarrow}^{t} near εF\varepsilon_{\mathrm{F}}, shown in Figure 5(g) and Figure 5(h), becomes larger than that in the inset of Figure 2(a), although the deformation of σst\sigma_{s}^{t} occurs due to the scattering rate −Im⁡Σsα-\operatorname{Im}\Sigma_{s}^{\alpha} shown in Figure 5(e) and Figure 5(f). This enhanced spectral conductivity leads to the reduction of ρ\rho.

Image not extracted

Figure 5. (Color online) Electronic states of n-type Fe2_2VAl with antisite defects obtained by the BPRAM. The left and right columns correspond to down-spin and up-spin quantities, respectively. (a), (b) The DOS per spin for total electrons (black solid line), valence electrons (red dashed line), and conduction electrons (blue dash-dotted line). (c), (d) The DOS per spin for VFe_{\mathrm{Fe}} and FeV_{\mathrm{V}} antisites. (e), (f) Imaginary part of the spin-dependent self-energy for valence (red solid line) and conduction (blue solid line) electrons. (g), (h) Zero-temperature spectral conductivity per spin for total (black solid line), valence (red dashed line), and conduction (blue dash-dotted line) electrons. The vertical lines indicate the Fermi energy εF\varepsilon_{\mathrm{F}}.

The magnitude of SS in Figure 4(b) is reduced in the presence of antisite defects at low temperatures compared with the case without antisites. This reduction is mainly attributed to the decreased slope of σst\sigma_{s}^{t} near εF\varepsilon_{\mathrm{F}}, as shown in Figure 5(g) and Figure 5(h), which is caused by antisite scattering. In particular, the conduction-band contribution σsc\sigma_{s}^{c} around ε=0.15\varepsilon=0.15 eV is suppressed owing to the presence of a peak in −Im⁡Σsc-\operatorname{Im}\Sigma_{s}^{c} at the same energy, leading to a gentler slope of σst\sigma_{s}^{t} near εF\varepsilon_{\mathrm{F}}. The valence-band contribution σsv\sigma_{s}^{v} around ε=−0.1\varepsilon=-0.1 eV is also suppressed due to a peak in −Im⁡Σsv-\operatorname{Im}\Sigma_{s}^{v}, resulting in a reduced hole contribution to SS in the vicinity of εF\varepsilon_{\mathrm{F}}. These effects contribute to the negative values of SS at low temperatures.

With increasing temperature above 300 K, the calculated SS in Figure 4(b) becomes positive, in contrast to the experimentally observed negative values for quenched Fe2_2VAl0.9_{0.9}Si0.1_{0.1},[20] shown by the blue dotted line. The origin of this discrepancy is unclear, but one possible explanation is a slight modification of γe−p\gamma_{\mathrm{e-p}} due to the presence of antisite defects in n-type Fe2_2VAl: a reduction of γe−p\gamma_{\mathrm{e-p}} in the conduction band would enhance the n-type carrier contribution at high temperatures.

Figure 5(b) and Figure 5(d) display the DOS of the antisite states, defined as Dsd=−1πIm⁡(Gsd,v+Gsd,c)D_{s}^{d}=-\frac{1}{\pi}\operatorname{Im}(G_{s}^{d,v}+G_{s}^{d,c}). Both VFe_{\mathrm{Fe}} and FeV_{\mathrm{V}} have spin polarization, since energies of their down- and up-spin states with εF\varepsilon_{\mathrm{F}} in between. Consequently, the total magnetic moment, given by the sum of the contributions from VFe_{\mathrm{Fe}} and FeV_{\mathrm{V}}, becomes approximately twice as large as that in Fe2_2VAl, where only VFe_{\mathrm{Fe}} antisites are spin-polarized,[25] when the antisite concentration is the same. This enhancement may account for the experimental observation that Si-substituted samples exhibit larger antisite-induced magnetic moments.[20]

Figure 6 shows the calculated ρ\rho and SS for p-type Fe2_2VAl with antisite defects. The magnitude of ρ\rho does not change significantly, although its slope is modified and its value at T<400T<400 K increases compared with the case without antisites. The magnitude of SS at T<400T<400 K is also enhanced in the presence of antisites. To confirm these predictions, systematic experiments that control the antisite concentration in p-type Fe2_2V1−x_{1-x}Tix_xAl are desired in the future.

Image not extracted

Figure 6. (Color online) Temperature dependence of (a) the calculated resistivity ρ\rho and (b) the Seebeck coefficient SS in the BPRAM for p-type Fe2_2VAl with antisite defects (red solid lines). For comparison, the results without antisite defects, identical to the red solid lines in Figure 3(b) and Figure 3(c), are shown by black dashed lines.

The corresponding electronic states are shown in Figure 7. The Fermi energy εF\varepsilon_{\mathrm{F}} lies inside the valence band, as shown in Figure 7(a). No spin polarization appears, since the DOS of VFe_{\mathrm{Fe}} and FeV_{\mathrm{V}} is located above εF\varepsilon_{\mathrm{F}} [see Figure 7(b)]. In Figure 7(d), we find that σsc\sigma_{s}^{c} near εF\varepsilon_{\mathrm{F}} is strongly suppressed due to the large −Im⁡Σsc-\operatorname{Im}\Sigma_{s}^{c} shown in Figure 7(c). This suppression is the origin of the enhancement of ρ\rho and SS at low temperatures in Figure 6.

Image not extracted

Figure 7. (Color online) Electronic states of p-type Fe2_2VAl with antisite defects obtained by the BPRAM. No spin polarization is present. (a) DOS per spin for the total electrons (black solid line), valence electrons (red dashed line), and conduction electrons (blue dash-dotted line). (b) DOS per spin for VFe_{\mathrm{Fe}}, which produces a shoulder, and for FeV_{\mathrm{V}}, which corresponds to a peak. (c) Imaginary part of the self-energy for the valence (red solid line) and conduction (blue solid line) electrons. (d) Zero-temperature spectral conductivity per spin for the total (black solid line), valence (red dashed line), and conduction (blue dash-dotted line) electrons. The vertical line represents the Fermi energy εF\varepsilon_{\mathrm{F}}.

Summary and discussions

Spin polarization of antisite defects gives rise to magnetic moments in Fe2_2VAl. In our previous work,[25] we found that VFe_{\mathrm{Fe}} is spin-polarized whereas FeV_{\mathrm{V}} is not in undoped Fe2_2VAl. In this paper, we show that both VFe_{\mathrm{Fe}} and FeV_{\mathrm{V}} are spin-polarized in n-type Fe2_2VAl, while neither of them is polarized in p-type Fe2_2VAl, as discussed in § 3. This difference originates from the relative positions of the antisite energy levels with respect to εF\varepsilon_{\mathrm{F}}, which depend on whether the system is n-type or p-type.

Figure 8 summarizes the schematic electronic states in Fe2_2VAl with antisite defects (the right side of the vertical energy axis). From this illustration, we can predict that antisite-induced magnetic moments decrease from n-type to p-type Fe2_2VAl, provided that the antisite concentration is the same. The recent experimental observation that Si-substituted samples exhibit larger antisite-induced magnetic moments compared with the undoped case[20] may reflect this behavior. It is also interesting to note that the magnetic moments in p-type Fe2_2V1−x_{1-x}Mnx_xAl decrease with increasing xx.[14] If the magnetic moment at x=0x = 0 originates from antisite defects, a possible source of the reduction may be the shift of εF\varepsilon_{\mathrm{F}} deeper into the valence band, as illustrated in Figure 8(c). Further systematic investigations of the effects of antisite defects in doped Fe2_2VAl are desired in the future.

Image not extracted

Figure 8. (Color online) Schematic illustration of the electronic states and the spectral conductivity in Fe2_2VAl with antisite defects. (a) Undoped Fe2_2VAl, (b) n-type, and (c) p-type Fe2_2VAl. The top of the valence band is set as the energy reference. The Fermi energy εF\varepsilon_{\mathrm{F}} is indicated by horizontal dashed lines. The left side of the vertical energy axis represents the spectral conductivity averaged over spins for the valence band, σv(ε)\sigma^{\mathrm{v}}(\varepsilon) (red solid curve), and for the conduction band, σc(ε)\sigma^{\mathrm{c}}(\varepsilon) (blue solid curve). The electronic states are shown on the right side of the vertical energy axis. The valence band is composed of Fe 3d orbitals, whereas the conduction band consists of V 3d orbitals. The DOS per spin (D↑D_{\uparrow} and D↓D_{\downarrow}) for the valence and conduction bands is shown by black solid lines. The antisite DOS for VFe_{\mathrm{Fe}} and FeV_{\mathrm{V}} is depicted by red and blue solid lines, respectively. In (a), the energy level of VFe_{\mathrm{Fe}}, denoted by εdv\varepsilon_{\mathrm{d}}^{\mathrm{v}}, appears below εF\varepsilon_{\mathrm{F}} for the ↑\uparrow-spin antisite state, while the ↓\downarrow-spin state is split upward by the on-site Coulomb interaction UdU_{\mathrm{d}}. Both spin states of FeV_{\mathrm{V}} appear above εF\varepsilon_{\mathrm{F}} at εdc\varepsilon_{\mathrm{d}}^{\mathrm{c}}. In (b), the spin states of FeV_{\mathrm{V}} are also split by UdU_{\mathrm{d}}, whereas in (c) no splitting occurs for either VFe_{\mathrm{Fe}} or FeV_{\mathrm{V}}.

In the left side of the vertical energy axis in Figure 8, the spectral conductivity for the valence (conduction) band, σv(c)(ε)=σ↑v(c)(ε)+σ↓v(c)(ε)\sigma^{\mathrm{v(c)}}(\varepsilon) = \sigma_{\uparrow}^{\mathrm{v(c)}}(\varepsilon) + \sigma_{\downarrow}^{\mathrm{v(c)}}(\varepsilon), is shown by red (blue) solid curves. These results highlight the interplay between thermoelectric and magnetic properties in Fe2_2VAl with antisite defects. In Figure 8(a), σv(ε)\sigma^{\mathrm{v}}(\varepsilon) near εF\varepsilon_{\mathrm{F}} is suppressed and flattened due to the spin-↑\uparrow level of the spin-polarized VFe_{\mathrm{Fe}} antisite. These changes reduce the hole contribution to SS, resulting in a negative Seebeck coefficient, consistent with the behavior reported for undoped Fe2_2VAl.[25] In n-type Fe2_2VAl, shown in Figure 8(b), the spin-↑\uparrow level of spin-polarized FeV_{\mathrm{V}} suppresses and flattens σc(ε)\sigma^{\mathrm{c}}(\varepsilon). Consequently, σv(ε)\sigma^{\mathrm{v}}(\varepsilon) and σc(ε)\sigma^{\mathrm{c}}(\varepsilon) take similar values near εF\varepsilon_{\mathrm{F}}, leading to a small Seebeck coefficient at low temperatures, as demonstrated in Figure 4(b). In p-type Fe2_2VAl, all antisite levels lie above εF\varepsilon_{\mathrm{F}}, and their influence on the spectral conductivity near εF\varepsilon_{\mathrm{F}} is weak, as shown in Figure 6(b). These close relationships between spectral conductivity and spin-polarized electronic states demonstrate the crucial role of an. antisite defects in controlling both thermoelectric and magnetic properties of Heusler-type alloys. Dynamical effects associated with fluctuations of antisite magnetic moments, which are not taken into account in the present study, may further contribute to the modification of SS.

Image not extracted

Figure 9. (Color online) Calculated power factor (PF) of undoped and doped Fe2_2VAl with (solid lines) and without (dashed lines) antisite defects. The black, red, and blue lines represent the results for undoped, nn-type, and pp-type Fe2_2VAl, respectively.

The power factor (PF), defined as S(T)2/ρ(T)S(T)^2/\rho(T), serves as a guideline for evaluating the energy conversion efficiency of thermoelectric materials. Figure 9 summarizes the calculated PF for undoped and doped Fe2_2VAl with and without antisite defects. In the absence of antisites, PF increases with carrier doping. This enhancement is attributed to both the decrease in ρ\rho and the increase in the absolute value of SS, as shown in Figure 2 and Figure 3. The PF of pp-type Fe2_2VAl is larger than that of nn-type Fe2_2VAl. This trend is opposite to experimental data.[11] This difference arises because the calculated SS is larger and smaller than the experimental values for pp-type and nn-type Fe2_2VAl, respectively.

Image not extracted

Figure 3. (Color online) Electronic states and transport properties in p-type Fe2_2VAl. (a) Total DOS per spin. The energy zero is set at the top of the valence band. The inset shows the zero-temperature total spectral conductivity per spin. The vertical line denotes the position of the Fermi energy εF\varepsilon_{\mathrm{F}}. (b) Temperature dependence of the resistivity ρ\rho. (c) Temperature dependence of Seebeck coefficient SS. The red lines represent calculated results. The black dashed and dotted lines correspond to experimental data for Fe2_2V1−x_{1-x}Tix_xAl with x=0.05x = 0.05 and x=0.2x = 0.2 in (b) and x=0.03x = 0.03 and x=0.1x = 0.1 in (c).[11] For comparison, the calculated results for Fe2_2VAl$[25] are shown by black dash-dotted lines in (b) and (c).

With antisites, the calculated PF of undoped Fe2_2VAl[25] decreases, in contrast to the experimentally observed large enhancement of PF in fully quenched Fe2_2VAl with large antisite concentration.[21] This discrepancy primarily arises from the small magnitude of the calculated SS, which reflects the low antisite concentration assumed in our model.[25] In addition, dynamical contributions from fluctuations of antisite magnetic moments may lead to a further enhancement of SS. In nn-type Fe2_2VAl, the introduction of antisite defects reduces the PF, which is consistent with experimental behavior.[20] In pp-type Fe2_2VAl, the PF exhibits nearly the same magnitude with and without antisites. This small change is attributed to the absence of spin-polarized antisites, which leads to weaker scattering of hole carriers near εF\varepsilon_{\mathrm{F}}.

In summary, we have investigated the thermoelectric properties and spin polarization associated with antisite defects in doped Fe2_2VAl. We have demonstrated that, in doped Fe2_2VAl, both the rigid-band shift of the Fermi energy εF\varepsilon_{\mathrm{F}} and the temperature-dependent scattering rate play crucial roles in understanding the temperature dependence of the Seebeck coefficient and the resistivity based on the Sommerfeld-Bethe relationship. In the presence of VFe_{\mathrm{Fe}} and FeV_{\mathrm{V}} antisite defects, the BPRAM with a shifted εF\varepsilon_{\mathrm{F}} yields spin polarization at both types of antisites in nn-type Fe2_2VAl, giving rise to a magnetic moment that is approximately twice as large as that of antisites in undoped Fe2_2VAl, when the antisite concentration is the same. At the same time, the Seebeck coefficient is reduced in the presence of antisites. These results are consistent with existing experimental observations. In contrast, in pp-type Fe2_2VAl, spin polarization of antisites diminishes because the Fermi energy lies deep inside the valence band and the antisite levels are located above εF\varepsilon_{\mathrm{F}}. These theoretical results demonstrate the impact of carrier substitution on the spin polarization of antisite defects in Fe2_2VAl and indicate the crucial role of antisites in realizing magneto-thermoelectric functionalities in Heusler-type alloys. Constructing a theoretical framework that takes into account the dynamical effects of antisite magnetic moments remains a challenge for future studies aiming to understand magneto-thermoelectric properties.

References

  1. [9]9) H. Kato, M. Kato, Y. Nishino, U. Mizutani, and S. Asano, J. Japan Inst. Metals 65, 652 (2001).
  2. [10]10) C. S. Lue, C. F. Chen, J. Y. Lin, Y. T. Yu, and Y. K. Kuo, Phys. Rev. B 75, 064204 (2007).
  3. [11]11) H. Matsuura, Y. Nishino, U. Mizutani, and S. Asano, J. Japan Inst. Metals, 66, 767 (2002).
  4. [12]12) A. Ślebarski, J. Goraus, J. Deniszczyk, and Ł. Skoczeń, J. Phys.: Condens. Matter 18, 10319 (2006).
  5. [14]14) R. Jha, N. Tsujii, F. Garmoudi, S. Khmelevskyi, E. Bauer, and T. Mori, J. Mater. Chem. C, 12, 8861 (2024).
  6. [15]15) F. Garmroudi, A. Riss, M. Parzer, N. Reumann, H. Müller, E. Bauer, S. Khmelevskyi, R. Podloucky, T. Mori, K. Tobita, Y. Katsura, and K. Kimura, Phys. Rev. B 103, 085202 (2021).
  7. [17]17) S. Jemima, A. Mani, A. Bharathi, N. Ravindran,and Y. Hariharan, J. Alloys Compd. 268, 183 (2001).
  8. [18]18) E. P. Amaldass, A. T. Satya, S. Sharma, K. Vinod, V. Srinivas, C. S. Sundar, and A. Bharathi, J. Alloys Compd. 648, 34 (2015).
  9. [19]19) N. Tsujii, A. Nishide, J. Hayakawa, and T. Mori, Sci. Adv. 5, eaat5935 (2019).
  10. [20]20) F. Garmroudi, M. Parzer, M. Knopf, A. Riss, H. Michor, A. V. Ruban, T. Mori, and E. Bauer, Phys. Rev. B 107, 014108 (2023).
  11. [21]21) F. Garmroudi, M. Parzer, A. Riss, A. V. Ruban, S. Khmelevskyi, M. Reticcioli, M. Knopf, H. Michor, A. Pustogow, T. Mori, and E. Bauer, Nat. Commun. 13, 3599 (2022).
  12. [23]23) F. X. Zhang, C. Feng, Z. Guan, H. Song, S. Nan, Y. Tang, H. Yang, K. Zeng, X. Yuan, M. Xiao, Z. Lu, Y. Zhang, D. Ma, and Y. Zhang, Acta Materi. 292, 121056 (2025).
  13. [24]24) P. W. Anderson, Phys. Rev. 124, 41 (1961).
  14. [25]25) T. Tohyama and H. Fukuyama, J. Phys. Soc. Jpn. 94, 094603 (2025).
  15. [28]28) M. Ogata and H. Fukuyama, J. Phys. Soc. Jpn. 88, 074703 (2019).
  16. [29]29) M. Matsubara, T. Yamamoto, and H. Fukuyama, J. Phys. Soc. Jpn. 94, 024601 (2025).

Paper details

vlnc:W_6ma5wqxj
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents